MiniMax image-01 API: 9 images, 1,500 characters, 24-hour URLs vs Sume
MiniMax image-01 takes a 1,500-character prompt, 1 to 9 images, 512-2048 px sizes and returns URLs that expire in 24 hours. Sume's Image API limits compared.

MiniMax's image-01 text-to-image API takes a prompt of up to 1,500 characters, returns 1 to 9 images per request, accepts width and height from 512 to 2048 pixels in multiples of 8, and returns URLs that expire after 24 hours. Sume's Image API takes up to 10 images per call, with lower per-model ceilings, and delivers results through the job's result artifacts.
The table lists each limit, and the sections below cover errors, references and what to change in a client that moves between the two.
What are image-01's request limits?
The reference lists image-01 as the only model value. The prompt is capped at 1,500 characters. aspect_ratio defaults to 1:1 and accepts 1:1, 16:9, 4:3, 3:2, 2:3, 3:4, 9:16 and 21:9. Width and height each run from 512 to 2048, must be divisible by 8, and the page says aspect_ratio takes priority when both are sent. n is 1 to 9, seed makes results reproducible, and prompt_optimizer defaults to false.
The response format is url or base64, defaulting to url, and the page notes a URL expires in 24 hours. The response carries image URLs or base64 strings, a success_count and failed_count in metadata, and a trace id. A status code of 0 means success.
| Item | MiniMax image-01 | Sume Image API |
|---|---|---|
| Prompt length | 1,500 characters | No single cap; read the model catalog |
| Images per call | n = 1 to 9 | n up to 10; per-model ceilings are lower |
| Sizing | aspect_ratio or width/height 512-2048, multiples of 8 | Normalized aspect_ratio and resolution tier, or image_size on some models |
| Reproducible output | seed | Read supported_parameters for the model |
| Prompt rewriting | prompt_optimizer, off by default | Not a shared field; model-specific |
| Result delivery | URL (24 hours) or base64 | Job result artifacts |
How do errors and references differ?
MiniMax returns its status inside the body, in base_resp, rather than as only an HTTP code. The reference lists examples: 1002 for rate limit, 1004 for failed authentication, 1008 for insufficient balance, 1026 for sensitive content detected, 2013 for invalid parameters and 2049 for an invalid API key. A client should read base_resp before assuming success.
For references, the guide describes subject_reference for consistent identity, limited to one image per request and typed as a character. The text-to-image schema excerpt does not list it, so read the image-to-image reference before relying on it. On Sume, references go in input_references as public HTTPS URLs, and a model whose descriptor is zero to zero rejects them.
How do you move a client between them?
Treat limits as data, not constants. On Sume, read the model's supported_parameters and the n range from the catalog before you send, and fetch the artifact URL from the job result instead of storing a link. On MiniMax, download within 24 hours or request base64.
curl https://api.sume.com/v1/images/models \
-H "Authorization: Bearer $SUME_API_KEY"When is each a good fit?
MiniMax image-01 fits when you want explicit pixel sizes inside 512 to 2048 and a seed on a single model, and when you already hold a MiniMax account. Sume fits when images are one input to a larger pipeline: the same job envelope, wallet and webhook as video and audio, and a catalog of several image models behind one request shape.
Check each vendor's current pricing page before you compare cost, since neither the MiniMax reference nor this page gives a per-image price.
- Need pixel-exact sizes and a seed: image-01 on MiniMax.
- Need many image models in one client: Sume's Image API.
- Need to keep results for more than a day: download them from either route and store them yourself.
- Need to avoid sensitive-content rejections: read the status code or reason before retrying.
Sources
Related posts
More in Developers
- MiniMax task statuses vs Sume job statuses: a mapping table
MiniMax reports queued, running, succeeded, failed, cancelled. Sume job status uses queued, processing, completed, failed, canceled. Map them correctly.
- Mirage Tesseract needs local files; Sume needs public HTTPS URLs
Mirage Tesseract runs on local files in an agent environment. Sume avatar and lip-sync inputs must be public HTTPS URLs. How to hand clips between the two.
- Music API has no duration field: steer length in the prompt
Sume's music router rejects duration and duration_seconds. Ask for a 30-second or 2-minute track in the prompt, with section timestamps, and verify.
- music_create over MCP: dry run, idempotency key and jobs_wait
How the hosted MCP music_create tool works: prompt inside payload, dry_run preview, a required idempotency_key, and jobs_wait in 55-second slices.
Written by Sume